Peregian Digital HubAI Adopted

Built 2026-06-21

Public AI agents as a learning engine: River at Shopify

A source note from the desk: synopsis, claims, relevance, caveats, and the original post preserved below for context.

Summary

The Shopify CEO describes River, an internal AI coding agent that lives in public Slack channels rather than private windows. The core insight is that public work with AI creates an unintentional learning engine: engineers watching other engineers use River adopt the same patterns the next day, and the agent itself improves as teams document what it should know. In the past month, 5,938 Shopify employees used River across 4,450 channels; roughly one in eight pull requests merged that week came from River. Because all interactions are searchable and visible, the company builds what Lütke calls a "Lehrwerkstatt"—a teaching workshop where learning spreads by proximity rather than formal training. Read the full piece at the original post.

Key Claims

  • Public-by-design AI agents become a learning engine because staff absorb patterns by watching others work with the agent, without formal training or curriculum.
  • River's merge rate climbed from 36% to 77% over two months without model retraining: teams noticed where it got stuck, fed back what it should have known, and the agent improved by absorbing each team's accumulated knowledge.
  • Private AI windows (ChatGPT, Claude, Cursor) lock apprentices out; whoever is at the keyboard learns, everyone else is locked out of the model's improvement loop.
  • Org speed is set by the slowest communication channel; searchable, public work with an agent spreads information faster than meetings, email, or private DMs and creates a record the next person finds.
  • The agent does not replace the mentor or apprentice; it makes the whole company an apprentice by putting experienced people's work on display so others can watch both human and AI learn.

Quotes

  • "Every conversation is therefore searchable. Anyone at Shopify can jump in."
  • "The risk is not that AI does the work. The risk is that AI does the work and we never learn from it."
  • "The best prompt patterns spread, knowledge spreads. The clever way one developer investigated a Slack permissions bug becomes the template for how everyone else investigates."